MISpheroID: a knowledgebase and transparency tool for minimum information in spheroid identity

Abstract

Spheroids are three-dimensional cellular models with widespread basic and translational application across academia and industry. However, methodological transparency and guidelines for spheroid research have not yet been established. The MISpheroID Consortium developed a crowdsourcing knowledgebase that assembles the experimental parameters of 3,058 published spheroid-related experiments. Interrogation of this knowledgebase identified heterogeneity in the methodological setup of spheroids. Empirical evaluation and interlaboratory validation of selected variations in spheroid methodology revealed diverse impacts on spheroid metrics. To facilitate interpretation, stimulate transparency and increase awareness, the Consortium defines the MISpheroID string, a minimum set of experimental parameters required to report spheroid research. Thus, MISpheroID combines a valuable resource and a tool for three-dimensional cellular models to mine experimental parameters and to improve reproducibility.

Fig. 1. Mapping the reporting topography in breast cancer spheroid research.

Binary heatmap showing the experimental parameters (rows, 51 of 98 parameters, selected for relevance) of each spheroid experiment (columns, n = 1,628). The heatmap is divided vertically into three sections of parameters (‘setup’, ‘characterization both microscopic and non-microscopic’ and ‘application’; indicated in blue, light and dark green, and red; and including 18, 21 + 7, and 5 parameters, respectively) and horizontally according to the year of publication. For each section, rows are sorted in descending order according to total number of reported experimental parameters. Parameters that were not reported in an experiment appear as a white space in its corresponding column. The reporting efficiency of each parameter is indicated as a percentage in the right column. EM, electron microscopy; ECM, extracellular matrix. Source data

Fig. 2. Reporting efficiency of experimental parameters in spheroids from different tumor types.

Spider plot visualization of the reporting percentage of cell line, culture medium, formation method and size in spheroids from different tumor types (clockwise). Axes represent the percentage of reporting efficiency. Source data

Fig. 3. Culture media-induced transcriptional variation.

a, Principal component analysis of gene expression profiles from spheroids of four cancer cell lines cultured in six different medium types. b, Heatmap of Z-scores of all MSigDB hallmark gene sets identified by GSEA to be significantly enriched among the differentially expressed genes across the culture medium types in each cell line. Medium types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Fig. 4. Culture media-induced heterogeneity in spheroid metrics across multiple cell types.

a, Spider plots of metrics from spheroids of indicated cell lines cultured in six different medium types. Axes represent the Z-score metrics of cell death, ATP content, L/G ratio, secreted protein signatures of angiogenesis and immune interaction, circularity, size and therapy response. A higher Z-score means a higher metric value. The left and middle columns indicate established cell lines; early passage and patient-derived sarcoma cultures are on the right. b, Cell death of HCT116 spheroids cultured in six different medium types evaluated at seven different laboratories (“sites”) in an interlaboratory experiment. At each site a higher ranking indicates a higher cell death. Each dot represents an evaluated spheroid. The colors indicate the medium type as in a. Source data

Fig. 5. Effect of formation methodology and spheroid size-induced heterogeneity on spheroid metrics across multiple cell types.

a, Spider plot of metrics from spheroids of indicated cell lines generated by the hanging drop (red) or liquid overlay (green) spheroid formation method. Axes represent the Z-score metrics of cell death, ATP content, secreted protein signatures of angiogenesis and immune interaction, circularity and size. A higher Z-score means a higher metric value. b, Violin plots representing the impact of spheroid size on cell death (upper panel) and ATP content (lower panel) metrics of HCT116 spheroids cultured in six different media. Each biological replicate has a different symbol (N ≥ 3), and each symbol is a technical replicate (n = 8). Triangles at the X axis represent increasing seeding cell number and consequently increasing spheroid size; for absolute size estimates see Supplementary Table 6. The Y axis represents log₂-transformed data, and all media are normalized to DMEM HG. The horizontal bar indicates the median. Statistical significance between the groups was determined with a one-way ANOVA and Tukey’s multiple comparison test. **P < 0.01, ***P < 0.001, ****P < 0.0001. Colors indicate medium type; media are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Fig. 6. Implementation of the MISpheroID knowledgebase.

This flowchart illustrates the application of MISpheroID.

Extended Data Fig. 1. Distribution of the applied culture medium to prepare spheroids of the estrogen-dependent MCF7 and T47D, and triple-negative MDAMB231 and 4T1 breast/mammary gland cancer cell lines.

Pie chart visualizing the proportion of culture media used to prepare MCF7, T47D, MDAMB231 and 4T1 spheroids. Source data

Extended Data Fig. 2. Distribution of the applied culture medium to prepare spheroids of the most reported cell line from non-breast tumors.

Pie chart visualizing the proportion of culture media used to prepare A549 (lung), HCT116 (colorectal), HEPG2 (liver), PANC1 (pancreas), SKOV3 (ovarium) and U87MG (brain). Source data

Extended Data Fig. 3. Comparison of the formulation of frequently reported culture media in spheroid research.

Heat map showing the formulation of culture media types used in the empirical MISpheroID study, in addition with other media types frequently reported in spheroid research practices including L15, HamF12, HamF12K and McCoy’s5A. Ranked according to decreasing nutrient richness from left to right. Nutrients that are not included in a medium type appear as a grey space in its corresponding column. Left column indicates nutrient type. Numerical value of nutrients in each medium is indicated in each cell. Values are concentrations, expressed in mM, unless indicated *=µM, # =nM, $=mg/l.

Extended Data Fig. 4. Quantitative presentation of the impact of heterogeneity in culture medium on cell death in spheroids.

Left, Z-score heatmaps and right, violin plots presenting the impact of six different media types on cell death in (a) 8 established cell lines and (b) 3 early passage, patient-derived sarcoma cultures. Biological replicates are indicated by a different symbol (N ≥ 3); each symbol is a technical replicate (n = 8). Y-axis represents log2-transformed data, all media types are normalized to DMEM HG. Horizontal bar indicates median. Statistical significance between the groups was determined with a one-way ANOVA and Tukey’s multiple comparison test. **p < 0.01, ***p < 0.001, ****p < 0.0001. Colors in violin plots present media type as in Fig. 3a; media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Extended Data Fig. 5. Image presentation of cell death, circularity and size in spheroids cultured in six different media types.

Representative microscopy images show ethidium homodimer I stained (red if cell is dead) spheroids of all evaluated cell lines and patient-derived sarcoma cells (cell types indicated on the left) cultured in different media types (indicated on the top) (scale bars 200 µm). Each experiment was repeated independently at least 3 times with 8 technical replicates per experiment, with similar results. Media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Intense staining in the spheroid center is indicative of necrotic core. Next to cell death, images give an indication of circularity and size. Source data

Extended Data Fig. 6. Quantitative presentation of the impact of heterogeneity in culture medium on ATP content in spheroids.

Left, Z-score heatmaps and right, violin plots presenting the impact of six different media types on ATP content in (a) 8 established cell lines and (b) 3 early passage, patient-derived sarcoma cultures. Biological replicates are indicated by a different symbol (N ≥ 3); each symbol is a technical replicate (n = 8). Y-axis represents log2-transformed data, all media types are normalized to DMEM HG. Horizontal bar indicates median. Statistical significance between the groups was determined with a one-way ANOVA and Tukey’s multiple comparison test. **p < 0.01, ***p < 0.001, ****p < 0.0001. Colors in violin plots present media type as in Fig. 3a; media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Extended Data Fig. 7. Quantitative presentation of the impact of heterogeneity in culture medium on ratio of glucose uptake to lactate secretion in spheroids.

Left, Z-score heatmap and right, scatter diagram presenting the impact of six different media types on ratio of glucose (G) uptake to lactate (L) secretion (L/G ratio) in 5 established cell lines (N ≥ 2, n = 4). Indicative trendlines are presented in the scatter diagram with a dotted line (higher slope means a higher ratio of lactate secretion to glucose uptake (for example y = 2x; for the consumption of one glucose molecule 2 lactate molecules are produced)). Error bars indicate SD. Colors in scatter diagram present media type as in Fig. 3a; media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Extended Data Fig. 8. Quantitative presentation of the impact of heterogeneity in culture medium on size of spheroids.

Left, Z-score heatmaps and right, violin plots presenting the impact of six different media types on spheroid size difference in (a) 8 established cell lines and (b) 3 early passage, patient-derived sarcoma cultures. Biological replicates are indicated by a different symbol (N ≥ 3); each symbol is a technical replicate (n = 8). Y-axis represents log2-transformed changes in size, all media types are normalized to DMEM HG. Horizontal bar indicates median. Statistical significance between the groups was determined with a one-way ANOVA and Tukey’s multiple comparison test. **p < 0.01, ***p < 0.001, ****p < 0.0001. Colors in violin plots present media type as in Fig. 3a; media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Extended Data Fig. 9. Response to radiotherapy of spheroids cultured in diverse media types.

Bar plots (a and c); and Z-score heatmaps (b and d) respectively indicate the impact of a single (a and b) 20 Gy and (c and d) 10 Gy fraction on ATP content of spheroids of indicated cell lines cultured in different media types. Biological replicates are indicated by a different symbol (N = 2; except HEPG2 N = 1); each symbol is a technical replicate (n = 8). Statistical significance between the groups was determined with a one-way ANOVA and Tukey’s multiple comparison test. **p < 0.01, ***p < 0.001, ****p < 0.0001. Horizontal bar indicates median, error bars indicate SD. Colors in bar plots present media type as in Fig. 3a; media types are ranked from higher nutrient (left) to lower nutrient (right) richness. Source data

Extended Data Fig. 10. Evaluation of the metrics cell death, circularity and size in an interlaboratory study.

(a) Correlation matrix of cell death ranking of HCT116 spheroids cultured in six different media types between different participating sites in the interlaboratory study (left) and between biological replicates in the study-initiating laboratory (right). In (b) and (c), Z-score heatmaps (left) and correlation matrices (right) present the impact of spheroid circularity (b) and size (c) of HCT116 spheroids cultured in six different media types evaluated across seven different sites in an interlaboratory setting. Media types in Z-score heatmaps are ranked from higher nutrient (left) to lower nutrient (right) richness. The median Spearman correlation across the entire dataset is respectively 0.56 for circularity and 0.57 for size. The larger spread in correlation coefficient of circularity across laboratories is due to subtle absolute differences with most probably limited biological significance. Source data